Performing inverse halftoning on a halftoned image includes smoothing the halftoned image using a two-dimensional filter to produce a smoothed image, detecting edge areas in the smoothed image, performing lowpass filtering on non-edge areas of the smoothed image, and generating an enhanced image comprised of the edge areas of the smoothed image and lowpass-filtered non-edge areas of the smoothed image.

Patent
   7228002
Priority
May 14 2001
Filed
May 14 2001
Issued
Jun 05 2007
Expiry
Aug 22 2024
Extension
1196 days
Assg.orig
Entity
Large
3
14
all paid
7. A method of enhancing an image, comprising:
smoothing the image using a sharply peaked filter to produce a smoothed image;
detecting an edge in the smoothed image;
performing lowpass filtering using a high frequency cutoff filter on the smoothed image to produce an enhanced image; and
detecting an edge in the enhanced image; and
applying a median filter to the enhanced image, wherein the median filter is designed to reduce artifacts on the enhanced image only on non-edge areas of the enhanced image as determined by the edge detection of the enhanced image.
1. A method of enhancing an image, comprising:
first, smoothing the image using a sharply peaked filter to produce a smoothed image;
detecting an edge in the smoothed image;
performing lowpass filtering on the smoothed image to produce an enhanced image, wherein lowpass filtering is performed using a high frequency cutoff filter only on non-edge areas of the smoothed image as determined by the edge detection; and
detecting an edge in the enhanced image; and
performing median filtering to the enhanced image only on non-edge areas of the enhanced image as determined by the edge detection of the enhanced image.
13. An article comprising a machine-readable medium that stores machine-executable, instructions for enhancing an image, the instructions causing a machine to:
first, smooth the image using a sharply peaked filter to produce a smoothed image;
detecting an edge in the smoothed image;
perform lowpass filtering using a high frequency cutoff filter on the smoothed image to produce an enhanced image; and
detecting an edge in the enhanced image; and
apply a median filter to the enhanced image, wherein the median filter is designed to reduce artifacts on the enhanced image only on non-edge areas of the enhanced image as determined by the edge detection of the enhanced image.
18. An apparatus for enhancing an image, comprising:
a memory that stores executable instructions; and
a processor that executes the instructions to:
first, smooth the image using a sharply peaked filter to produce a smoothed image;
detect an edge in the smoothed image; and
perform lowpass filtering using a high frequency cutoff filter on the smoothed image to produce an enhanced image,
detecting an edge in the enhanced image; and
wherein the processor executes instructions to apply a median filter only on non-edge areas of the enhanced image as determined by the edge detection of the enhanced image, and wherein the median filter is designed to reduce artifacts on the enhanced image.
14. An apparatus for enhancing an image, comprising:
a memory that stores executable instructions; and
a processor that executes the instructions to:
first, smooth the image using a sharply peaked filter to produce a smoothed image;
detect an edge in the smoothed image;
perform lowpass filtering using a high frequency cutoff filter on the smoothed image to produce an enhanced image, wherein lowpass filtering is performed only on non-edge areas of the smoothed image as determined by the edge detection; and
detecting an edge in the enhanced image; and
perform median filtering to the enhanced image only on non-edge areas of the enhanced image as determined by the edge detection of the enhanced image.
9. An article comprising a machine-readable medium that stores machine-executable instructions for enhancing an image, the instructions causing a machine to:
first, smooth the image using a sharply peaked filter to produce a smoothed image;
detect edge areas in the smoothed image;
perform lowpass filtering using a high frequency cutoff filter on the smoothed image to produce an enhanced image, wherein lowpass filtering is performed only on non-edge areas of the smoothed image as determined by the edge detection; and
detecting an edge in the enhanced image; and
perform median filtering to the enhanced image only on non-edge areas of the enhanced image as determined by the edge detection of the enhanced image.
2. The method of claim 1, wherein smoothing comprises:
applying a two-dimensional filter to a pixel in the image;
storing a pixel processed by the two-dimensional filter in the smoothed image; and
repeating storing and applying for one or more other pixels in the image.
3. The method of claim 1, wherein performing lowpass filtering comprises:
applying a one-dimensional filter to a pixel in the smoothed image;
storing a pixel processed by the one-dimensional filter in the enhanced image; and
repeating storing and applying for one or more other pixels in the smoothed image.
4. The method of claim 1, wherein detecting the edge comprises applying an edge filter to the smoothed image.
5. The method as recited in claim 4, wherein detecting an edge further comprises comparing a predetermined threshold with results of edge filtering, and wherein edge values determined by the edge filtering that exceed the threshold are ignored during lowpass filtering.
6. The method as recited in claim 1, wherein the method of enhancing an image is performed in one pass.
8. The method of claim 7, wherein detecting the edge comprises applying an edge filter to the smoothed image.
10. The article of claim 9, wherein smoothing comprises:
applying a two-dimensional filter to a pixel in the image;
storing a pixel processed by the two-dimensional filter in the smoothed image; and
repeating storing and applying for one or more other pixels in the image.
11. The article of claim 9, wherein performing lowpass filtering comprises:
applying a one-dimensional filter to a pixel in the smoothed image;
storing a pixel processed by the one-dimensional filter in the enhanced image; and
repeating storing and applying for one or more other pixels in the smoothed image.
12. The article of claim 9, wherein detecting the edge comprises applying an edge filter to the smoothed image.
15. The apparatus of claim 14, wherein smoothing comprises:
applying a two-dimensional filter to a pixel in the image;
storing a pixel processed by the two-dimensional filter in the smoothed image; and
repeating storing and applying for one or more other pixels in the image.
16. The apparatus of claim 14, wherein performing lowpass filtering comprises:
applying a one-dimensional filter to a pixel in the smoothed image;
storing a pixel processed by the one-dimensional filter in the enhanced image; and
repeating storing and applying for one or more other pixels in the smoothed image.
17. The apparatus of claim 14, wherein detecting the edge comprises applying an edge filter to the smoothed image.
19. The apparatus of claim 18, wherein lowpass filtering is performed only on non-edge areas of the smoothed image, and wherein detecting the edge comprises applying an edge filter to the smoothed image.

This invention relates to inverse halftoning.

Digital halftoning recreates an original shaded image using light and dark pixels. Varying the density and placement of the light and dark pixels in the resulting digitally-halftoned image approximates shading gradations in the original image. Inverse halftoning is a process of enhancing a digitally-halftoned image so that it more closely resembles the original shaded image.

FIG. 1 is a flowchart showing an inverse-halftoning process according to one embodiment of the invention.

FIG. 2 is a view of pixels in images processed by the inverse halftoning process.

FIG. 3 is a block diagram of computer hardware on which the inverse halftoning process may be implemented.

Referring to FIG. 1, a process 5 is shown for performing inverse halftoning on a halftoned image. Process 5 is basically a two-part process, with enhancements. In FIG. 1, the enhancements are shown by dotted lines.

In the basic two-part process, process 5 lightly smoothes (101) the halftoned image and then performs (103) lowpass filtering on the halftoned image. To lightly smooth the halftoned image, process 5 applies a two-dimensional (e.g., N×N, where N>1) filter to each pixel in the image.

In one embodiment, a strongly-peaked 3×3 filter is used to perform the light smoothing. Examples of filters to produce different shading effects are (A) and (B) below.

1 6 1 1 2 1
6 8 6 * (1/36) 2 8 2 * (1/20)
1 6 1 1 2 1
(A) (B)

A filter is applied to an image as follows. The filter matrix is overlaid on a block of N×N (e.g., 3×3) pixels. Each numerical value in the filter matrix is multiplied by a corresponding underlying pixel value. The resulting products are added together and multiplied by the reciprocal of the sum of the filter values (e.g., ( 1/36) and ( 1/20) above). The resulting value is stored as a pixel in an intermediary smoothed image. The pixel in the intermediary smoothed image is at the same location in the smoothed image as the center pixel in the block of N×N pixels that were filtered in the original image.

To process other pixels, the filter matrix is moved to an adjacent pixel in the original image. The filter matrix may be moved horizontally or vertically. Referring to FIG. 2, as each pixel 6 in an original image 7 is processed, the resulting filtered pixel is stored at a corresponding location 8 in a smoothed image 9. The resulting smoothed image, or portion thereof, is lowpass filtered to produce an enhanced image. Lowpass filtering is performed in order to reduce artifacts in the image. Examples of artifacts include unwanted spots or image aberrations.

Process 5 performs (103) lowpass filtering using a one-dimensional lowpass filter having a relatively sharp high-frequency cutoff. Examples of well-known lowpass filters that have sharp high-frequency cutoffs, and that may be used, include Hamming and Kaiser filters. In this embodiment, a one-dimensional filter with N (N≧1) taps is used; however, a two-dimensional filter, whose dimensions in some, but not all, instances will be N×N, may be used to perform the lowpass filtering. The lowpass filter may have any number of “Itaps”. e.g., 11, 7, 3, etc., and may be weighted by a sinc function to vary filter width. The lowpass filter is moved over the smoothed image, and filters the smoothed image, in much the same way as the 3×3 filter described above.

However, unlike the case of filtering with a 3×3 two-dimensional matrix, filtering with a one-dimensional matrix is done by making two passes, one horizontal and one vertical, over the image. First, the lowpass matrix is overlaid along either a block of row pixels and moved horizontally or a block of column pixels and moved vertically to create an intermediate image. On the second pass, the filter is applied to the intermediate image along the direction not applied in the first pass. For example, a filter having the matrix values

The value is assigned to a pixel in the enhanced image that has the same location in the enhanced image as the pixel below the center pixel of the lowpass filter in the smoothed image. For example, as shown in FIG. 2, N-tap lowpass filter is applied to pixels 12, 14 and 16 of smoothed image 9. The resulting filtered value is assigned to pixel 20 in enhanced image 22. Pixel 20 in enhanced image 22 is at the same location as pixel 14 in smoothed image 9, which is at the center of the lowpass filter.

To process other pixels in the smoothed image, the lowpass filter is moved to an adjacent pixel in the smoothed image. That adjacent pixel is then processed using the two-pass filtering described above. As each pixel in the smoothed image is processed, the resulting filtered pixel value is stored at a corresponding location in the enhanced image. By repeating this process for plural pixels in the smoothed image, process 5 generates (104) an enhanced, inverse-halftoned, version of the smoothed image.

Additional features may be incorporated into process 5. These features are shown in dotted lines on FIG. 1. For example, process 5 may detect (102) edges in the smoothed image and ignore the edges during lowpass filtering (103), i.e., only lowpass-filter the non-edge areas of the smoothed image. Thus, in this case, the enhanced image is comprised of the edge areas of the smoothed image and lowpass-filtered non-edge areas of the smoothed image. One reason for ignoring the edges during lowpass filtering is that lowpass filtering tends to blur the image. To ensure that the edges remain relatively defined, process 5 may detect those edges and perform lowpass filtering selectively, such that the edges are not lowpass filtered.

To detect edges in the smoothed image, process 5 applies an edge filter to the smoothed image. The edge filter may be an N×N filter, which processes pixels in substantially the same manner as the N×N filter described above. That is, the filter matrix is overlaid on a block of N×N (e.g., 3×3) pixels. Each numerical value in the filter is multiplied by a corresponding underlying pixel value. The resulting products are added together to obtain a value. In this case, unlike above, the value is not multiplied by the reciprocal of the sum of the pixel values. If the resulting value has a large magnitude, that is an indication of an edge in the image.

One example of an edge detection filter matrix that may be used in this embodiment is as follows:

−1 0 1
−1 0 1
−1 0 1

Applying this edge detection filter to a flat surface of a pixel would result in a near zero value, since the pixels underneath the first column (−1's) would cancel the pixels underneath the third column (1's). If an edge is present at the center column, the value resulting from applying the filter over the center pixel and its adjacent eight pixels would result in a non-zero magnitude because values along one center are different from other adjacent values. The higher this magnitude, the more pronounced the edge.

Process 5 compares the value that results from applying the edge detection filter to a threshold value. Edges whose values exceed the threshold are ignored during lowpass filtering. This threshold value is set beforehand and is set, as desired, to ignore some edges and account for others. The higher that the threshold value is set, the more edges the process 5 ignores during lowpass filtering (103). That is, if the threshold is set relatively high, only highly-pronounced edges will not be subjected to lowpass filtering, whereas if the threshold is set relatively low, relatively less-pronounced edges will also not be subjected to lowpass filtering.

Process 5 may also apply (105) a median filter to the resulting enhanced image. The median filter is designed to reduce artifacts, such as spots or other aberrations, in the enhanced image. The median filter may also be an N×N filter, which is applied in the manner described above, to further smooth the enhanced image. The median filter may be applied to non-edge portions of the enhanced image only, if desired. In this case, an edge detection filter of the type described above is used to detect edges in the enhanced image prior to applying the median filter to the enhanced image.

FIG. 3 shows a computer 28 on which process 5 may be implemented. Computer 28 includes a processor 30, a memory 32, and a storage medium 34 (see view 36). Storage medium 34 stores data 38 for one or more images and machine-executable instructions 40 that are executed by processor 30 out of memory 32 to perform inverse halftoning process 5.

Although a personal computer is shown in FIG. 3, process 5 is not limited to use with the hardware and software of FIG. 3. It may find applicability in any computing or processing environment. Process 5 may be implemented in hardware, software, or a combination of the two. Process 5 may be implemented in computer programs executing on programmable computers or other machines that each include a processor, a storage medium readable by the processor (including volatile and non-volatile memory and/or storage components), at least one input device, and one or more output devices. Program code may be applied to data entered using an input device (e.g., a mouse or keyboard) to perform process 5 and to generate output information.

Each such program may be implemented in a high level procedural or object-oriented programming language to communicate with a computer system. However, the programs can be implemented in assembly or machine language. The language may be a compiled or an interpreted language.

Each computer program may be stored on a storage medium/article (e.g., CD-ROM, hard disk, or magnetic diskette) that is readable by a general or special purpose programmable computer for configuring and operating the computer when the storage medium or device is read by the computer to perform process 5. Process 5 may also be implemented as a machine-readable storage medium, configured with a computer program, where, upon execution, instructions in the computer program cause a machine to operate in accordance with process 5.

The invention is not limited to the specific embodiments described above. For example, a Sobel gradient operator may be used for edge detection instead of, or in addition to, the N×N matrix described above. The invention is not limited to the specific filters described herein, to their numerical values, or to their dimensions. The invention can be used to perform inverse halftoning on images received over a network or on any other types of halftoned images.

Other embodiments not described herein are also within the scope of the following claims.

Macy, Jr., William W.

Patent Priority Assignee Title
7522220, Mar 30 2005 Samsung Electronics Co., Ltd. Dual-channel adaptive 2D noise reduction for video signals
9215345, Mar 26 2013 Xerox Corporation Method and system for inverse halftoning utilizing inverse projection of predicted errors
9521296, Aug 27 2014 Xerox Corporation Inverse halftoning using inverse projection of predicted errors for multi-bit images
Patent Priority Assignee Title
5027078, Oct 10 1989 Xerox Corporation Unscreening of stored digital halftone images by logic filtering
5166810, Aug 30 1989 FUJI XEROX CO , LTD , A CORP OF JAPAN Image quality control system for an image processing system
5323247, Dec 04 1990 ROCHESTER, UNIVERSITY OF Method and apparatus for halftoning and inverse halftoning and the transmission of such images
5333064, Apr 30 1991 Creo IL LTD Apparatus & method for descreening
5343309, Aug 12 1992 Xerox Corporation Image processing system and method employing adaptive filtering to provide improved reconstruction of continuous tone images from halftone images including those without a screen structure
5506699, Apr 27 1993 HEWLETT-PACKARD DEVELOPMENT COMPANY, L P Method and apparatus for converting a halftone image into a continuous tone image
5798846, May 28 1996 HEWLETT-PACKARD DEVELOPMENT COMPANY, L P Apparatus and method for selectively processing a scanned image
5850294, Dec 18 1995 THE CHASE MANHATTAN BANK, AS COLLATERAL AGENT Method and apparatus for post-processing images
5852475, Jun 06 1995 Cisco Technology, Inc Transform artifact reduction process
6201613, Jul 22 1998 Xerox Corporation Automatic image enhancement of halftone and continuous tone images
6229578, Dec 08 1997 Mobil Oil Corporation Edge-detection based noise removal algorithm
6621595, Nov 01 2000 HEWLETT-PACKARD DEVELOPMENT COMPANY, L P System and method for enhancing scanned document images for color printing
6947178, Feb 26 2001 Ricoh Company, LTD De-screening halftones using sigma filters
JP3259193,
//
Executed onAssignorAssigneeConveyanceFrameReelDoc
May 08 2001MACY, WILLIAM W , JR Intel CorporationASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS 0118150036 pdf
May 14 2001Intel Corporation(assignment on the face of the patent)
Date Maintenance Fee Events
Jan 10 2011REM: Maintenance Fee Reminder Mailed.
Mar 10 2011M1551: Payment of Maintenance Fee, 4th Year, Large Entity.
Mar 10 2011M1554: Surcharge for Late Payment, Large Entity.
Nov 19 2014M1552: Payment of Maintenance Fee, 8th Year, Large Entity.
Nov 22 2018M1553: Payment of Maintenance Fee, 12th Year, Large Entity.


Date Maintenance Schedule
Jun 05 20104 years fee payment window open
Dec 05 20106 months grace period start (w surcharge)
Jun 05 2011patent expiry (for year 4)
Jun 05 20132 years to revive unintentionally abandoned end. (for year 4)
Jun 05 20148 years fee payment window open
Dec 05 20146 months grace period start (w surcharge)
Jun 05 2015patent expiry (for year 8)
Jun 05 20172 years to revive unintentionally abandoned end. (for year 8)
Jun 05 201812 years fee payment window open
Dec 05 20186 months grace period start (w surcharge)
Jun 05 2019patent expiry (for year 12)
Jun 05 20212 years to revive unintentionally abandoned end. (for year 12)